Blog on AJhttps://ajsquestions.github.io/blog/Recent content in Blog on AJhttps://ajsquestions.github.io/AJpicture.jpghttps://ajsquestions.github.io/AJpicture.jpgHugo -- gohugo.ioenMon, 06 Apr 2026 00:00:00 +0000Why ML Models Disagree About Stock Returnshttps://ajsquestions.github.io/blog/why-ml-models-disagree-about-stock-returns/Mon, 06 Apr 2026 00:00:00 +0000https://ajsquestions.github.io/blog/why-ml-models-disagree-about-stock-returns/When neural networks and gradient boosting look at the same stocks, they see different things. Here's why that disagreement is geometrically structured — and what it means for building better ensembles.What Regularization Actually Does to Your Portfoliohttps://ajsquestions.github.io/blog/what-regularization-actually-does-to-your-portfolio/Fri, 20 Mar 2026 00:00:00 +0000https://ajsquestions.github.io/blog/what-regularization-actually-does-to-your-portfolio/Ridge regression doesn't just shrink coefficients — it encodes a belief about which directions in return space matter. Here's the geometric intuition.What I Learned Building a Football Prediction Systemhttps://ajsquestions.github.io/blog/building-a-football-prediction-system/Sun, 15 Feb 2026 00:00:00 +0000https://ajsquestions.github.io/blog/building-a-football-prediction-system/FootieM8 covers 25+ leagues across 30 seasons. The models work. The markets are harder.